This sounds like a good strategy to me. It's expensive to iterate on big trucks per unit. Iterate on the cars. More cars. More focus. Overlap.
Blank check is also a terrible policy and making Waymo execs get their own funding puts their skin in the game more. Sounds like they're just weeding out the slackers.
I thought that automated trucking was by far the best possible fit for self-driving vehicles. Dealing with the enormous amount of insanity that consumers experience day to day: predominantly residential, stray children/pets/cross walkers, aberrant weather, construction, etc feels insurmountable without actual intelligence.
In contrast, a professional trucking company could have far more reliable interstate-only medium-haul routes. Truck could operate only when conditions are favorable, otherwise human would take that day's route. Auto-Truck would drive from one hub point to another and park itself until a human was available to complete the delivery to the real endpoint. Not sure how well the systems deal with night, but you could even prioritize Auto-Truck to run after hours when there is less traffic and potential complications.
I've never driven a truck, but I get the impression that even simple interstate driving still has lots of possibilities for most of the issues you mentioned but also the vehicle is the size and weight of a small building. I was surprised to learn for instance that you can fill a standard truck up so full of potatoes that it can come in over the legal weight limit and then subject the driver to a fine (I also note that in the US, the driver is subject to the fine, but in Europe it's the company that owns the truck who is liable which really sums up a lot of labor issues pretty nicely).
I suspect that for simple, regular routes, a train is going to be really effective already. Trucks are used because they are much more versatile.
In regards to violations holding the striver instead of the company liable, I see it exactly the opposite way you do. It empowers individual workers to refuse to create a dangerous situation. If the company is the one liable, the worker can wash their hands of the situation and happily make a dangerous situation on the road. The company won’t care and can treat it as a cost of doing business/lean on insurance risk pooling because things almost never go wrong making the risk cheap. If no individual is responsible, the only people who lose are the worker and other road users.
I would think having an amount of weight that could easily crush a car behind you traveling at highway speeds would create lots of incentive for the driver to play by the rules - it's actually one of the more dangerous jobs out there for that reason. Companies can buy more trucks and drivers though.
I used to think that too but if you're interested take a look at this debate between two experts in the field, they're both skeptical of trucking being first to go fully autonomous: https://www.youtube.com/watch?v=hbnS6tWiw88
Professional trucking companies tend to be small/medium are getting squeezed pretty hard by the larger businesses on either end. Any logistics company with a large enough fleet is probably investigating this themselves.
Why would you need to iterate anything in the design of a heavy truck? You are writing software that guides some hunk of metal from point A to point B given sensors I,J,K, GPS, and a database about the road. You don't need to make a special truck, you explicitly want a "kit" to convert existing ICE trucks.
This is not Tesla making an electric tractor.
Waymo/Uber/Tesla (especially Tesla) have really been a pox on the #1 thing I want: get me from city A to city B without me blankly staring at two lines and slightly adjusting a steering wheel 1% every minute for hours on end. Instead everyone is chasing some use case that e-scooters and e-bikes will likely be superior to in 80% of use cases.
It's a problem (urban taxis / urban cars) that city planning and evolution is explicitly moving away from. It's far harder to do, and probably needs convergent infrastructure investment/buildout that would be better pioneered with automated long haul driving.
I'm really hoping what this really means is that the long haul problem is being addressed better with other less sexy industry-aligned startups or projects.
"Waymo's focus on ride-hailing makes some sense. The reliability requirements for ride-hailing are much lower than trucking, making it a more lenient business. If you have a truck full of cargo, it's a major issue if something goes wrong and it can't reach its destination on time. The truck routes are many hours long over long distances and usually have some kind of delivery time attached. Your self-driving hardware and software has to work perfectly during all that. Ride-hailing is way easier. Trips are usually measured in minutes and in a localized area where you can easily dispatch support people if something goes wrong. Because the app is a central point of customer bookings, you can easily pause and resume accepting customers anytime. That makes it easy to shut down the fleet to deal with technical difficulties or bad weather. You can also rigidly control your service area and accept or decline trips on a whim. Everyone can just use Lyft instead."
Most of this smells like bullshit to me.
If your technology can't hire extremely well-mapped, generally well-maintained highways, it can't possibly handle cities. Saying that you expect to fail and kick out the passengers and tell them to use some other service is in no way a usable service.
I think Waymo is toast, and they're trying for a soft landing and some kind of pivot.
This self driving over hyped tech only works in good weather and is political suicide to go after trucking with it since so many high paying blue collar jobs will be consumed.
It only makes sense if one doesn't know much about self-driving. What they quote and refer to as "bullshit" is very reasonable, these systems do break down because they can't handle every possible situation on their own. They do not have human intelligence.
With robotaxis what they currently do is human drivers go out and rescue the vehicle. You can't do that with a truck, it's not economical. This will be an issue for years to come, we're probably going to see more remote assist but unlikely for an 80,000 pounds killing machine driving at highway speeds.
Which brings us to another point: Regulation and risk perception. Cruise having some hickups and maybe being involved into some minor fender benders is one thing. Politicians and some of the public in the respective cities are already concerned. But a fully loaded self driving semi being involved in a crash and potentially killing some people could be the autonomous version of 9/11. That's a lot of risk, the licenses might be revoked and then their investment is toast.
